Case Study: Effects of In-Game NBA Fatigue
The game of basketball is fast-paced, action-packed, and physical. Athletes exert a tremendous amount of energy and get worn down every night they suit up, let alone throughout a whole 82 game season. Maximizing lineups and player efficiencies is a difficult thing to do, which motivated Joey Devirgilio and me to explore this topic – in-game NBA fatigue – in a case study for the UMass Sports Analytics Club’s 2017 Case Competition.
I wanted to share some of our findings with you.
The goal of this case study is to discover and explore the trends of fatigue in the NBA by comparing the fluctuations in players’ performances throughout games. Looking at key signifiers that suggest fatigue has occurred will give us a better understanding of how relevant it is throughout a game.
It was found that there are indeed signs that show lethargy towards the end of games. Comparing the first quarter with the fourth quarter provides the greatest discrepancies, so we will focus most on these times. To have a full slate of data, we will use stats from the 2015-16 season.
We wanted to focus on six factors: Dunks, time of possessions, shot selections, team pace, overplaying players, and player preferences.
This was a very basic factor that we initially used to see if the effect of in-game fatigue was noticeable enough to dive deeper into out research.
A general assumption that goes with measuring weariness has been dunks. Many people within the basketball analytics community believe that as players get more tired, the frequency of dunks lowers due to the explosiveness and athleticism required in dunking. So, we looked at how many dunks were made and attempted in the first quarter versus the fourth quarter.
In the 4th quarter, the amount of dunks attempted decreased by 12.31% with over 300 less being attempted. It’d be fairly safe to assume fatigue played a role in this decrease in dunks. Other factors may have affected these results as well, including defenses playing tighter in the 4th quarter of games, but fatigue without a doubt was a huge factor. Players being tired would be less inclined to drive to the basket and instead may be more inclined to settle for jump shots when they are fatigued.
Time of Possessions
Perhaps one of the best indicators of fatigue is how early in a possession teams are shooting. If they are getting shots off quickly, it means they are either forcing turnovers, having a lot of opportunities in transition, pushing the pace, or (most likely) all of the above. In order for any of this to be happening, players would have to be moving around the court very quickly or harassing opponents on defense to force turnovers, both of which require a decent amount of energy and stamina. Under this logic, teams would have more scores and attempts early in the shot clock in the 1st quarter as opposed to the 4th.
One way to measure the average time of possessions is using filters to look at how often teams shoot based on time left on the shot clock. The timing of possessions are classified as follows:
Very early shot: 22-18 seconds on shot clock
Early shot: 18-15 seconds on shot clock
Average shot: 14-7 seconds on shot clock
Late shot: 7-4 seconds on shot clock
Very late shot: 4-0 seconds on shot clock
The score of games is a major component that could interfere with our findings because if a team has a lead with only a few minutes remaining, they will be looking to run out the clock and slow down the pace on offense in the 4th quarter.
As expected, the frequency of late and very late shots increased from the first quarter to the fourth quarter. “Very early shots” were taken less often, falling from 14.94% in the first to 12.95% in the fourth, and “early shots” fell from 17.90% to 14.85%, as indicated in bold font below:
In other words, all early shots (in green) were taken 5.04% less often in the fourth quarter, while late shots (in red) increased by 5.38%.
The frequency of “average shots” stays consistent, but there is still a trend towards a longer average time of possessions in the fourth quarter:
To go along with the drop off in dunks, observing the change in reliance on outside shooting and isolation, along as the quality of those shots, helps understand [the impacts of] in-game fatigue. Settling for more contested shots and isolation plays points towards fatigue. To find when and how often this was occurring we looked at multiple factors: Pull-up versus catch-and-shoot shots, touch time (not included here), and defender distance.
First, we exported the first and fourth quarter data for pull-up and catch-and-shoot frequencies and compared the changes. Secondly, the amount of touch time on each shot can show when isolations or individual-playmaking are taking place. To measure this, we categorized the touch time of each shot into <2 seconds, 4-6 seconds, and 6+ seconds. Lastly, defender distance can demonstrate how often teams are getting open shots. We analyzed this by filtering team shooting into the following categories when shots were at least 10 feet from the rim:
Very tight shots: 0-2 feet
Tight shots: 2-4 feet
Open shots: 4-6 feet
Wide open: 6+ feet
The results show that all three shooting factors support each other and the idea of fatigue affecting shot types and efficiency. Looking first and the type of shots, there is a 2.84% increase in pull-up shots in the fourth quarter and a 0.61% decrease in catch-and-shoot shots:
The quality of shots also changes throughout a game. Less open shots suggest that players cannot create offense as well in the fourth quarter and settle for contested shots. The frequency of contested shots increased by 2.48% in the fourth quarter:
Now that could just mean that a) defenses have adjusted, b) defenses are playing tighter defense since it is the final quarter or c) offenses are settling for bad shots due to a decrease in off-ball movement and activity This trend nonetheless supports the in-game fatigue argument.
Shot touch times is the third factor that shows the adjustments in shots from the first to the fourth quarter, and our results fall right in the trend that teams take worse shots towards the end of games. There is a 3.68% heavier reliance on isolation and individual play in the fourth (teams have the ball for 6+ seconds before shooting 3.68% more often):
Similar to possession time, the score could potentially interfere with these results. However, when looking at games won and lost in the fourth quarter, the data still favored a slower pace with longer touch times on shots. The frequency for 6+ second touch time shots still increases from the 10.09% in the first quarter (shown above) to 12.43% in lost fourth-quarter games.
There are a wide variety of offensive philosophies in the NBA and this lends itself to some teams playing at a much faster pace, and going in transition more than others. The question at hand is whether these fast paced teams hurt themselves by making their players more fatigued. Using the Pace statistic we looked at top 10 and bottom 10 teams in the NBA and their field goal percentages in the first and fourth quarter. The difference between these two percentages were calculated to see if teams saw a drop off in their percentages as the game went along. The average of the difference in percentages for the Top 10 and Bottom 10 teams was also calculated to see how they stacked up against one another.
The results displayed an interesting conclusion different from what our hypothesis expected. For the ten fastest teams most but not all saw a drop in their field goal percentage from the first to fourth quarter. Two of these teams were the 76ers and Suns though, whose rise in percentage could be attributed to their horrible record and frequency of blowout games where starters weren’t present in most of the the fourth quarter:
When the ten slowest teams were analyzed, more teams saw a drop in field goal percentage. This was somewhat surprising because in theory if the teams were playing at a slower pace their player’s would be less fatigued and less affected by it in the fourth quarter. One interesting exception is the Spurs who’s field goals percentage rose the highest out of all teams that were tracked.
These results did not support our hypothesis. However, it is interesting to note that the Spurs had the largest spike in field goal percentage in the league from the first quarter to the fourth. Being known as a team who rests players and closely follow players’ nightly minutes, perhaps this is something that the rest of the league to take into account.
A fairly logical idea is that if fatigue plays a factor in player performance than players should not play a substantial amount of minutes as it will limit their effectiveness. We wanted to see if player’s field goal percentage got worse if they played more minutes than usual. To test this we decided to compare player’s field goal percentages when they play more than 40 minutes in a game (a huge amount) vs. when they played between 30 and 40 minutes (a relatively normal amount for a starter).
We compiled the stats for all of the players who played >40 minutes at least 8 times so that we had a decent chunk of data to compare. Then compared their field goal percentages for when they played >40 minutes and when they played between 30 and 40 minutes. The difference in these percentages was calculated and a T-Test was done to see if the data was significant.
When the data was compiled what was seen was varied results in terms of how players were affected by playing more minutes. Some players, like Kevin Durant, were more effective if they played fewer minutes. These players, who can be observed in the green colored boxes, had field goal percentages that were higher in the 30-40 minute range. Others, like James Harden, were the opposite. These players, noted in red, saw a drop in their field goal percentages when they played fewer minutes:
It must be noted there are a variety of other factors that played a role in these results including how some players may have only been left in the game for 40+ minutes when they were playing well.
Looking at these stats it is reasonable to assume that minutes played does not play a huge role in a player’s effectiveness and that other factors are more important than this one.
Varying Player Trends
JJ Redick and Avery Bradley are known to be two of the hottest shooters out of the gates of games. In 2015-16, Redick made 43 shots in the opening two minutes of the first quarter, 13 more than the next closest, and Bradley made the eighth most. This is valuable information to know, but what perhaps may be even more crucial to observe is when they start to cool off.
To address this question, we gathered player shooting for every minute of every quarter to compare trends in field goal percentage. Some players take some time to heat up before hitting their stride, while others need no time at all. Some players can only play efficiently for a small stretch of time, while others can perform at a high level for a long time.
Going back to the original example, Redick and Bradley both see significant drop-offs in field goal percentage in the first quarter, but at different times. After 10 minutes of 49.6% shooting, Bradley shot 25.8% in the closing two minutes of the quarter. Redick’s steady shooting of 53.5% plummets down to 23% at the eight-minute mark.
The results are as expected – players’ shooting is eventually negatively affected by fatigue over an extended period of consecutive time on the court and differs depending on the player, their role, and their nature.
Player Trends Example: Harrison Barnes
Harrison Barnes is a prime example of these findings. Here is Barnes’ shots made and field goal percentages through the first quarter last year:
As you can see, it took Barnes some time to get involved in the Warrior’s offense. His most effective and most common shooting came right around the five-minute mark. If he was kept on the court after that point, his production would fall off a cliff. These trends are something that we can observe for any player in the league and could prove to be a valuable tool for coaches.
The effects of in-game fatigue on NBA players is varied depending on what part of the game you are looking at. But as a whole the effects are important and should not be ignored when hoping to reach optimal efficiency out of your players.
The best way to look at this data is on a player by player basis. If this is done coaches can maximize their player’s effectiveness by using smart substitution methods that go along with the data on their players. This data in some cases can also be used as a scouting technique for teams trying to figure out other teams’ substitution patterns. If a player averages 60 shots in minutes 12-8 of the second quarter, and averages only 10 in minutes 8-4 it is pretty reasonable to assume what time he will be in the game. This data can be used in a variety of ways such as the one above, but no matter what it is reasonable to say it is not intelligent to ignore fatigue data in NBA players.
Follow Joeseph Devirgilio on Twitter: @Joeyd599
Photo via the Flickr Creative Commons, thanks to Matthew Addie for the shot of Barnes and company.